Spin diffusion results

The proton spin-diffusion decay curves for the rigid (crystalline) phase and the build-up curve for the soft (amorphous) component of the PBT/PET blend with and without GF are depicted in Figures 6.11 and 12, respectively. The magnetization flows from the rigid into the mobile domains reaching the equilibrium state after 50 ms for both blends.

Based on spin-diffusion curves, the samples with GF (Fig. 6.12) show lower equilibrium magnetization values for the amorphous fractions and higher values for the crystalline phases compared to the blend sample without GF (Fig. 6.11). This could be related to the presence of GF, which is mostly found in the mobile fractions. Chains which are close to the GF have strongly reduced mobility and behave thus like rigid chains contributing to the selected rigid fraction through the DQ filter.

The proton spin-diffusivities of the rigid and amorphous phases for the PBT/PET blends with and without GF were determined with equations 2.3 and 2.4 (Fig. 6.13). The amorphous fractions are more affected by the addition of GF while the spin-diffusivities for the crystalline domains are almost identical for both blends.

Figure 6.12: Depth-dependent proton spin-diffusion decays and build-up curves for the rigid (crystalline) and soft (amorphous) fractions of the PBT/PET blend with GF: 1) surface layer, 2) layer in the middle of the sample, and 3) other surface layer.

Figure 6.13: Depth-dependent spin-diffusion coefficients for the phase domains of the PBT/PET blends with and without GF.

Domain-size determination

The Domain sizes for the blend samples with and without GF were determined based on spin-diffusion experiments and upon employing equation 2.8 (Fig. 6.14). The sizes of the crystalline domains for the PBT/PET blend without GF show a maximum in the middle of the sample, which might result from temperature gradient across the sample upon processing conditions and subsequent cooling.

Figure 6.14: Depth-dependent domains size for the phase domains of the PBT/PET samples with and without GF.

For the sample with GF the domain size of the amorphous fraction is lower and the estimated domain sizes for the crystalline fraction are higher at both surfaces but lower in the middle of the sample as compared to those of the samples without GF. As a result, the average crystalline domain size for the blend with GF is 7.3 nm and slightly higher than that of the samples without GF with a crystalline domain size of 6.9 nm. This could be related to the reduced chain mobility of the amorphous phase near the GF. As a result, these chains are attributed to the fraction of the crystalline phase, leading to the increase in the domain size compared to the samples without glass fiber. For the same reason, the domain sizes of the amorphous phase of the samples with GF are lower than those without GF. The amorphous fraction of the blend samples with GF shows largely constant values, as the GF are well distributed in this fraction. In addition to the processing conditions, the nucleation effect of GF might also be responsible for the variations in the domains size. However it is difficult to distinguish the contribution of each effect based on the current study. The calculations for different dimensionalities [42] of the phase domains are presented in table 6.1, where the lamellar morphology is the most pronounced one for the PET and PBT phase domains [138].

Table 6.1: Extracted domain sizes for different morphologies.

Dimensionality PET/PBT PET/PBT + GF

dc [nm] da [nm] dc [nm] da [nm]

⃰ lamellar (ε = 1) 6.9 3.7 7.3 3.1

cylindrical (ε = 2) 13.6 7.3 14.6 6.3

spherical (ε = 3) 20.5 11 21.9 9.4

⃰ PET and PBT crystallize in a triclinic crystal structure leading to a lamellar morphology. Upon cooling from the melt, the crystallization of PET and PBT first involves the formation of nuclei and their subsequent growth. The latter is in the form of lamellae radiating outward from the nucleus by chain folding perpendicular to the direction of the growth [138].

6.4 Conclusions

Proton low-field relaxation measurements on PBT/PET blends using a Profile NMR- MOUSE allowed distinguishing two phase domains: a rigid phase and a mobile one. Depth-dependent relaxation measurements proved that the blend sample without GF showed a maximum T2eff for the soft fraction towards the middle of the plate whereas,

the surface layers showed almost the same values. This higher chain mobility can be explained with the increased transesterification rate towards the middle of the plate due to the production method and cooling process. However the rigid phase showed largely constant T2eff. Furthermore, the rigid fraction, which relates to the crystallinity, is also

increased towards the middle of the plate due to the lower cooling rate towards the depth. In contrast, the addition of GF results in a uniform morphology. This can result from the lower temperature gradient through the sample due to the enhanced thermal conductivity and reduction in transesterification in the presence of GF.

Furthermore, a methodology was developed to estimate the GF distribution with the help of low-field NMR measurements and by determining the relative amplitudes. The results showed a GF variation of about 12 % for PET-PBT blend samples with respect to the reference value of 50 % GF in the compound.

Line-shape deconvolution of 13_{C CPMAS spectra allowed the quantification of the}

crystalline and amorphous fractions as well as the PET/PBT ratio. Upon estimating the PBT/PET ratio, it was observed that the investigated blend samples behave as a gradient polymer, whereby the PBT content is dominant across the sample with around a 60 % contribution.

A spin-diffusion analysis recorded that the blend samples form a two-component system consisting of the crystalline and amorphous fractions. By a line-shape analysis of the obtained 1_{H spectra, the spin-diffusivities of the different morphological fractions}

were determined. The spin-diffusivities of the amorphous domains were lower for samples with GF, which might be related to a reduction in the chain mobilities due to the presence of GF. Furthermore, the domains size was estimated for the different morphological phases and various dimensionalities. The domain sizes for the crystalline fractions of the samples without GF showed a maximum in the middle of the sample, which might be related to the processing conditions. However, for the blend sample with GF, the average domain size was higher for the crystalline fraction and lower for the amorphous domains as compared to the samples without GF. This could be related to the reduced chain mobility of the amorphous phase in the vicinity of the GF.

Chapter 7

7 Non-destructive determination of

rubber content in rubber-modified

asphalt

7.1 Introduction and motivation

In the last recent decades, many efforts have been generated in order to improve the long-term performance and cost effectiveness of asphalt pavements. Polymer modification of bitumen (PMB) is one well-established procedure for improving the performance of asphalt pavements [139-140]. In particular, modification of bitumen with thermoplastic polymers and elastomers such as polyethylene (PE), polypropylene (PP), ethylene-vinyl acetate (EVA), ethylene-butyl acrylate (EBA), styrene-butadiene-styrene (SBS), styrene-isoprene-styrene (SIS), and styrene-ethylene/butylene-styrene (SEBS) have been reported [140]. Although all of these polymers improve bitumen properties to different degrees, their drawbacks limiting the future development of bitumen polymer modification can include high cost, low ageing resistance, and poor storage stability of polymer modified bitumen (PMB) [139-140].

Another approach to asphalt modification is the use of recycled tire rubber or vulcanized crumb rubber to modify asphalt binder and mixtures. The tendency towards using this approach is now increasing in the paving industry due to the prominent benefits and improvements such as improved resistance to fatigue cracking, improved aging and oxidation resistance, and lower pavement maintenance costs [141-142].

Figure 7.1: Rubber-modified asphalt produced by the wet process (left) and the dry process (right) [141].

Furthermore, this practice can reduce environmental pollution by avoiding waste tire disposal and burning.

Crumb rubber is produced by two major technologies: ambient mechanical grinding which is the most commonly used process and cryogenic grinding which is the more expensive process but generates smoother and smaller crumbs [143]. The primary source of feedstock is waste tires from autos and trucks. The ground crumb rubber is blended into the asphalt mixture at varying percentages in order to create a rubberized asphalt. There are two methods of adding the crumb rubber to the asphalt mixture: the “wet process” and the “dry process” (Fig. 7.1). In the wet process, crumb rubber is mixed with the bitumen at elevated temperature using some type of shearing blending. The binder/rubber mixture is allowed to “cook” or “digest” by keeping the rubber/binder mixture at high temperatures for 20 minutes to an hour or more. This allows the rubber to absorb the lighter ends of the binder and both swell and soften. The industry routinely adds other chemicals to the binder (like extenders, polymers, etc.) to produce a crumb rubber-modified asphalt binder. It is important to note that vulcanized rubber cannot melt or engage in material chemical exchanges with the binder, so the grains of rubber will retain their form and must be kept in suspension while the binder is being used in production [141].

In the dry process, recycled tire crumb rubber is blended with heated stone aggregate before mixing with bitumen. This process produces rubber/binder/aggregate amalgam where the rubber is randomly distributed throughout the mix and cannot separate. During and immediately after production, the crumb rubber absorbs the lighter ends of the binder, swells and softens. The dry process also produces a crumb rubber- modified asphalt mix that performs comparably with wet process rubber in the field [141- 142].

A broad range of physical, mechanical and rheological techniques are used to investigate the properties of asphalt mixtures and binders. Among these, mechanical testing is a common method used to evaluate the mixture performance in the terms of the stiffness modulus, permanent deformation and fatigue resistance [141]. Furthermore, the quantification of the rubber fraction along with the control of the rubber distribution in the asphalt mixture are key factors in the quality control of rubber- modified asphalt. While various investigations have been reported on the mixing processes, characterization and properties of crumb rubber-modified asphalt mixtures, there is no analytical study for investigating and monitoring the rubber fraction in the finished asphalt mixtures in a non-destructive fashion.

Nuclear Magnetic Resonance (NMR) is a well-known analytical tool in chemistry for molecular analysis and in medicine for diagnostic imaging. NMR machines are fitted with proper magnets, which typically are super-conducting to provide high field strength for high sensitivity. During the last years considerable progress has been achieved with the development of compact NMR machines employing permanent magnets, which provide sufficient sensitivity and robustness for nondestructive characterization in applications outdoors and on the factory floor [4]. In most cases, samples are drawn and placed inside the magnet where the field is strongest and most homogeneous. But there are also stray-field NMR sensors, like the NMR-MOUSE (MObile Universal Surface Explorer), which can be placed on an object to measure the NMR signal from a region of the object covered by the sensor. Such measurements are truly non- destructive and are applicable to electrically insulating and non-magnetic objects of any size as long as they are rich in hydrogen. The measurement signal is collected from a

sensitive slice a few millimeters away from the surface outside the sensor. Different surface regions of the object as well as depth profiles can be recorded by simply shifting the position of this sensitive slice through the sample [22].

This study reports the first use of the NMR-MOUSE to quantify the rubber fraction in rubber-modified asphalt. A new NMR methodology has been developed in the laboratory based on calibration curves obtained from reference samples. This is accomplished by measuring the transverse NMR relaxation signal and analyzing the relative amplitudes and variations in the molecular dynamics of the soft and rigid domains of asphalt samples. Furthermore, for actual field test a modified version of potable NMR-MOUSE set-up was used. The raw data from the field measurements need to be further inserted in the calibration model to estimate the rubber fraction in the field and study the field variations in the terms of the rubber distribution.

7.2 Experimental section

Five different asphalt samples with known rubber fractions were prepared under controlled laboratory conditions and were used as reference samples for NMR-MOUSE measurement (Table 7.1). The NMR experiments were performed with the Profile NMR- MOUSE PM5 by Magritek GmbH, Germany, with a proton resonance frequency of 17.1 MHz. The transverse magnetization decays of the samples were measured with a Carr– Purcell–Meiboom–Gill (CPMG) sequence from a 10 × 10 × 0.1 mm3 _{sensitive slice at 5}

mm depth into each sample using an echo time of 50 µs and depth interval of 200 µm. The decays were analyzed by fitting the experimental data to a double exponential function,

Figure 7.2: Experimental set-up for measuring the asphalt samples with a PM5 NMR- MOUSE. The NMR-MOUSE is the black block underneath the black plate on which the rubber cylinder rests. Its sensitive slice is located inside the asphalt cylinder. To acquire depth profiles, the NMR-MOUSE is lowered from measurement to measurement in a step-wise fashion, so that the distance between NMR-MOUSE and test object increases and the sensitive slice move outward towards the bottom of the object.

Table 7.1: Reference samples and corresponding rubber fractions

Samples ⃰ Rubber fraction (%)

1 4

2 12

3 8

4 16

5 0

⃰ Rubber fractions are the contents by the total binder mass.

The effective relaxation times T2eff are related to the molecular mobilities in each

fraction. The relative fractions of the relaxation components Ai_/(Arigid _{+ A}soft_{) represent}

the relative numbers of hydrogen atoms (molar fractions) of the asphalt components with different molecular mobilities. For reasons of sensitivity, the relative amplitude of components was approximated by the sum of the first 8 echoes. The NMR experiments were conducted at room temperature and above with the help of a climate chamber (Fig. 7.2). The later was used to design the calibration curve for the actual measurements in the fields.

A representative relaxation signal of rubber-modified asphalt measured with the NMR-MOUSE shows a discrete signal decay. Such a relaxation signal is acquired for each position of the sensitive volume inside the sample and subsequently analysed for estimating the rubber fraction using our laboratory procedure.

Figure 7.3: Representative relaxation decay of an asphalt sample at 40° C.

For the field measurements on roads, a modified version of portable NMR-MOUSE from Magritek GmbH Aachen was used. The field set-up includes the PM5 NMR- MOUSE, a KEA spectrometer, a car battery, a voltage converter and a laptop to run the relaxation measurement and save the data (Fig. 7.4). The NMR-MOUSE sensor is laterally displaced to scan different spots in one location. The raw data from the field measurements are fed into a calibration model to estimate the rubber fraction in the field and to assess the field variations in the terms of the rubber distribution.

Figure 7.5: T2eff (a) and fractions (b) of the rigid and soft components of asphalt sample 1 at 40° C. 0 1000 2000 3000 4000 5000 0 10 20 30 40 50 60 70 80 90 100 c o n c e n tr a ti o n [ % ] depth [mm] soft fraction hard fraction 0 1000 2000 3000 4000 5000 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 T 2 e ff [ m s ] depth [mm] soft fraction hard fraction a) b)

7.3 Results and discussion

The aim of the laboratory study was first to sort the received samples from low to high rubber content and secondly to estimate the rubber fraction in each asphalt mixture. The former was followed by applying the 1_{H relaxation measurements while for the}

determination of rubber fraction a calibration curve was designed based on the NMR relaxation measurements on the known reference samples at different temperatures.

7.3.1

H relaxation measurements at elevated temperature

A relaxation signal (Fig. 7.3) is acquired for each position of the sensitive volume inside the sample and subsequently analyzed for relaxation times and signal component amplitudes following eqn. (7.1). The soft fraction is attributed to the rubber fraction. The effective relaxation times T2eff and the fractions of the rigid and soft components of the

asphalt samples have been determined as a function of depth into the asphalt cylinder at 40 C (Fig. 7.5). The temperature was elevated from room temperature to improve the signal-to-noise ratio of the measurement and to investigate the capability of the NMR-MOUSE in differentiating the various samples with different rubber contents.

Figure 7.6: Depth profiles of asphalt sample 1, at 40° C calculated as the sums of the first 8, 32 and 64 data points (top to bottom).

0 1000 2000 3000 4000 5000 5 10 15 20 25 30 35 40 45 re la ti v e a m p lit u d e depth [mm]

profiles and to estimate concentrations was also explored. Instead of fitting the experimental relaxation decays (Fig. 7.3) with eqn. (7.1) each data point in a depth profile was derived from a sum of the first data points of the NMR relaxation signal acquired at each depth. Depending on the number of data points added, the signal-to-

noise ratio improves, but the data sum becomes more and more a mixture of component amplitude and relaxation time the more data points are added (Fig. 7.6).

The two approaches to analyse the NMR relaxation signals of the asphalt samples, i. e. fitting eqn. (7.1) for a component analysis and calculating the sum of the first data points were compared in an effort to estimate the rubber fraction (Fig. 7.7).

Either procedure of estimating the rubber fraction shows similar trends but fails to produce absolute values. The shortcoming of the component analysis is that the initial part of the experimentally determined signal amplitude is lost in the dead-time of the instrument, so that the rigid fraction is underestimated and the soft fraction is overestimated. The benefit of the signal sum is that it is largely determined by the signal from the soft fraction which decays slowly, while that of the rigid fraction decays rapidly. As a result, the data sum is better suited to estimate the rubber fraction. However, in order to calculate the absolute rubber fraction the resultant NMR data and the values of data sum need to be calibrated with known concentration data.

Figure 7.7: Comparison of NMR relaxation data for the 5 asphalt samples in terms of depth averages. a) T2eff of the soft domains from a component analysis of the relaxation

signals measured at 40° C. b) Rubber fractions of soft domains estimated from a component analysis of the relaxation signals measured at 40° C. c) Sum of the first 8 data points measured at 40° C. d) Sum of the first 8 data points measured at room temperature.